Programming Summary
Jose Dixon
June 9, 2022
knitr::opts_chunk$set(warning = FALSE, message = FALSE, error = TRUE)library(tidyverse)
library(reshape)
library(ggplot2)
library(modelr)
library(tinytex)
options(na.action = na.warn)Hadley Wickman Intro to Data Science This is the website for “R for Data Science”.https://r4ds.had.co.nz
getwd()## [1] "C:/Users/z3696/Documents/Document-Classification/classifier/Output"table <- read.csv("~/Document-Classification/classifier/Output/Table.csv")
head(table)## Year Sampling Technique Classifier Precision Recall
## 1 2010 Imbalanced N/A Naive Bayes 74.49 16.70
## 2 2010 Imbalanced N/A Logistic Reg 72.82 21.18
## 3 2010 Imbalanced N/A XGBoost 12.66 75.05
## 4 2010 Imbalanced N/A DecisionTree 59.00 22.77
## 5 2010 Imbalanced N/A Random Forest 22.65 69.45
## 6 2010 Undersampling NearMiss Naive Bayes 76.67 35.38precision = table[, 5]
recall = table[, 6]
classifier = table[, 4]
sampling = table[, 2]
technique = table[, 3]
year = table[, 1]Chapter 3 Data Visualization
ggplot(data = table) +
geom_point(mapping = aes(x = precision, y = recall))
ggplot(data = table) +
geom_point(mapping = aes(x = precision, y = recall))
ggplot(data = table) +
geom_point(mapping = aes(x = precision, y = recall))
ggplot(data = table) +
geom_point(mapping = aes(x = precision, y = recall, color = technique))
ggplot(data = table) +
geom_point(mapping = aes(x = precision, y = recall, size = technique))
# Left
ggplot(data = table) +
geom_point(mapping = aes(x = precision, y = recall, alpha = technique))
# Right
ggplot(data = table) +
geom_point(mapping = aes(x = precision, y = recall, shape = technique))
ggplot(data = table) +
geom_point(mapping = aes(x = precision, y = recall), color = "blue")
ggplot(data = table) +
geom_point(mapping = aes(x = precision, y = recall)) + facet_wrap(~ Technique, nrow = 2)
ggplot(data = table) +
geom_point(mapping = aes(x = precision, y = recall)) +
facet_grid(Sampling ~ Technique)
ggplot(table, aes(precision, recall)) + geom_boxplot()
ggplot(data = table) +
geom_point(mapping = aes(x = precision, y = recall)) +
geom_smooth(mapping = aes(x = precision, y = recall))
ggplot(data = table) +
stat_summary(
mapping = aes(x = precision, y = recall),
fun.min = min,
fun.max = max,
fun = median
)
# Problematic Code
# ggplot(table, aes(x = precision, y = recall)) +
# geom_point(size = 2, colour = "grey30") +
# geom_abline(
# aes(intercept = a1, slope = a2, colour = -dist),
# data = table(models, rank(dist) <= 10)
# )
# Error in x[!nas] : object of type 'closure' is not subsettableggplot(data = table) +
geom_boxplot(mapping = aes(x=precision, y=recall, alpha = classifier))
ggplot(data = table) +
geom_count(mapping = aes(x = precision, y = recall, alpha = classifier))
x <- precision
y <- recall
ggplot(table) +
geom_histogram(mapping = aes(x = y), binwidth = 0.1)
ggplot(data = table) +
geom_point(mapping = aes(x = precision, y = recall))
Chapter 7 EDA
ggplot(data = table) +
geom_bar(mapping = aes(x = classifier))
table %>%
count(cut_width(Precision, 0.5))## cut_width(Precision, 0.5) n
## 1 [0.25,0.75] 1
## 2 (8.25,8.75] 2
## 3 (9.25,9.75] 1
## 4 (9.75,10.2] 1
## 5 (10.2,10.8] 1
## 6 (10.8,11.2] 1
## 7 (11.8,12.2] 1
## 8 (12.2,12.8] 1
## 9 (12.8,13.2] 2
## 10 (18.2,18.8] 1
## 11 (21.8,22.2] 3
## 12 (22.2,22.8] 5
## 13 (22.8,23.2] 3
## 14 (23.2,23.8] 4
## 15 (23.8,24.2] 2
## 16 (24.2,24.8] 3
## 17 (24.8,25.2] 2
## 18 (25.2,25.8] 2
## 19 (25.8,26.2] 1
## 20 (26.2,26.8] 1
## 21 (26.8,27.2] 1
## 22 (28.2,28.8] 3
## 23 (28.8,29.2] 2
## 24 (29.8,30.2] 1
## 25 (30.2,30.8] 2
## 26 (30.8,31.2] 2
## 27 (32.2,32.8] 1
## 28 (48.2,48.8] 2
## 29 (48.8,49.2] 1
## 30 (49.2,49.8] 1
## 31 (51.2,51.8] 1
## 32 (54.2,54.8] 1
## 33 (54.8,55.2] 1
## 34 (55.2,55.8] 1
## 35 (55.8,56.2] 3
## 36 (56.2,56.8] 4
## 37 (56.8,57.2] 3
## 38 (57.2,57.8] 7
## 39 (57.8,58.2] 9
## 40 (58.2,58.8] 4
## 41 (58.8,59.2] 10
## 42 (59.2,59.8] 6
## 43 (59.8,60.2] 2
## 44 (60.2,60.8] 2
## 45 (60.8,61.2] 2
## 46 (61.8,62.2] 2
## 47 (62.2,62.8] 2
## 48 (63.8,64.2] 1
## 49 (65.8,66.2] 2
## 50 (67.2,67.8] 1
## 51 (68.2,68.8] 3
## 52 (69.8,70.2] 1
## 53 (70.2,70.8] 1
## 54 (70.8,71.2] 3
## 55 (71.2,71.8] 2
## 56 (71.8,72.2] 1
## 57 (72.2,72.8] 4
## 58 (72.8,73.2] 2
## 59 (73.2,73.8] 4
## 60 (73.8,74.2] 1
## 61 (74.2,74.8] 4
## 62 (74.8,75.2] 3
## 63 (75.2,75.8] 3
## 64 (75.8,76.2] 2
## 65 (76.2,76.8] 4
## 66 (77.2,77.8] 5
## 67 (77.8,78.2] 2
## 68 (78.2,78.8] 2
## 69 (78.8,79.2] 4
## 70 (79.2,79.8] 3
## 71 (79.8,80.2] 6
## 72 (80.2,80.8] 1
## 73 (80.8,81.2] 3
## 74 (81.8,82.2] 1
## 75 (82.2,82.8] 3
## 76 (82.8,83.2] 4
## 77 (83.8,84.2] 2
## 78 (84.2,84.8] 1
## 79 (84.8,85.2] 1
## 80 (85.2,85.8] 1
## 81 (85.8,86.2] 1
## 82 (86.8,87.2] 2
## 83 (87.8,88.2] 1
## 84 (88.8,89.2] 1
## 85 (90.2,90.8] 2
## 86 (90.8,91.2] 1
## 87 (91.2,91.8] 2
## 88 (91.8,92.2] 2
## 89 (92.2,92.8] 1
## 90 (92.8,93.2] 2
## 91 (93.2,93.8] 2
## 92 (93.8,94.2] 2
## 93 (94.2,94.8] 5
## 94 (94.8,95.2] 2
## 95 (95.2,95.8] 2
## 96 (95.8,96.2] 1
## 97 (97.2,97.8] 1table %>%
count(Classifier)## Classifier n
## 1 DecisionTree 55
## 2 Logistic Reg 55
## 3 Naive Bayes 55
## 4 Random Forest 6
## 5 XGBoost 54table %>%
count(cut_width(Recall, 0.5))## cut_width(Recall, 0.5) n
## 1 [0.25,0.75] 1
## 2 (1.25,1.75] 2
## 3 (3.75,4.25] 2
## 4 (4.75,5.25] 1
## 5 (6.25,6.75] 1
## 6 (7.75,8.25] 1
## 7 (8.75,9.25] 1
## 8 (10.2,10.8] 1
## 9 (10.8,11.2] 1
## 10 (11.2,11.8] 1
## 11 (11.8,12.2] 1
## 12 (12.2,12.8] 1
## 13 (13.2,13.8] 1
## 14 (14.2,14.8] 2
## 15 (15.8,16.2] 1
## 16 (16.2,16.8] 1
## 17 (18.2,18.8] 2
## 18 (19.2,19.8] 1
## 19 (20.2,20.8] 1
## 20 (20.8,21.2] 2
## 21 (21.2,21.8] 2
## 22 (21.8,22.2] 2
## 23 (22.8,23.2] 2
## 24 (23.8,24.2] 1
## 25 (28.2,28.8] 1
## 26 (29.2,29.8] 1
## 27 (29.8,30.2] 1
## 28 (30.8,31.2] 4
## 29 (31.8,32.2] 1
## 30 (32.2,32.8] 1
## 31 (32.8,33.2] 2
## 32 (34.8,35.2] 3
## 33 (35.2,35.8] 1
## 34 (35.8,36.2] 5
## 35 (36.2,36.8] 2
## 36 (36.8,37.2] 5
## 37 (37.2,37.8] 1
## 38 (37.8,38.2] 3
## 39 (38.8,39.2] 3
## 40 (39.2,39.8] 1
## 41 (39.8,40.2] 4
## 42 (40.2,40.8] 6
## 43 (40.8,41.2] 2
## 44 (41.2,41.8] 2
## 45 (41.8,42.2] 5
## 46 (42.2,42.8] 4
## 47 (43.2,43.8] 7
## 48 (43.8,44.2] 4
## 49 (44.2,44.8] 3
## 50 (44.8,45.2] 1
## 51 (45.2,45.8] 2
## 52 (45.8,46.2] 1
## 53 (46.2,46.8] 4
## 54 (46.8,47.2] 2
## 55 (47.8,48.2] 1
## 56 (48.2,48.8] 1
## 57 (48.8,49.2] 1
## 58 (53.2,53.8] 1
## 59 (53.8,54.2] 4
## 60 (54.2,54.8] 1
## 61 (54.8,55.2] 1
## 62 (55.2,55.8] 1
## 63 (64.2,64.8] 1
## 64 (69.2,69.8] 2
## 65 (70.2,70.8] 1
## 66 (71.2,71.8] 1
## 67 (73.2,73.8] 1
## 68 (73.8,74.2] 1
## 69 (74.2,74.8] 2
## 70 (74.8,75.2] 1
## 71 (75.2,75.8] 1
## 72 (76.8,77.2] 4
## 73 (77.8,78.2] 1
## 74 (78.2,78.8] 1
## 75 (78.8,79.2] 3
## 76 (79.2,79.8] 1
## 77 (79.8,80.2] 1
## 78 (80.2,80.8] 1
## 79 (80.8,81.2] 2
## 80 (81.8,82.2] 2
## 81 (82.2,82.8] 2
## 82 (82.8,83.2] 2
## 83 (83.8,84.2] 5
## 84 (84.2,84.8] 1
## 85 (85.2,85.8] 2
## 86 (86.2,86.8] 3
## 87 (86.8,87.2] 1
## 88 (87.2,87.8] 3
## 89 (87.8,88.2] 5
## 90 (88.2,88.8] 8
## 91 (88.8,89.2] 5
## 92 (89.2,89.8] 4
## 93 (89.8,90.2] 4
## 94 (90.2,90.8] 6
## 95 (90.8,91.2] 2
## 96 (91.2,91.8] 4
## 97 (91.8,92.2] 2
## 98 (92.8,93.2] 4
## 99 (93.2,93.8] 3
## 100 (94.2,94.8] 1
## 101 (95.8,96.2] 2
## 102 (96.8,97.2] 1
## 103 (97.2,97.8] 3smaller <- table %>%
filter(Classifier > 60)ggplot(data = smaller, mapping = aes(x = Precision)) +
geom_histogram(binwidth = 0.1)
ggplot(data = smaller, mapping = aes(x = Precision)) +
geom_histogram(binwidth = 0.1)
ggplot(data = smaller, mapping = aes(x = Recall, colour = classifier)) +
geom_freqpoly(binwidth = 0.1)
ggplot(data = smaller, mapping = aes(x = Recall)) +
geom_histogram(binwidth = 0.01)
ggplot(data = smaller, mapping = aes(x = Precision, colour = classifier)) +
geom_freqpoly(binwidth = 0.1)
ggplot(data = smaller, mapping = aes(x = Precision)) +
geom_histogram(binwidth = 0.01)
ggplot(data = smaller, mapping = aes(x = Precision)) +
geom_histogram(binwidth = 0.25)
ggplot(table) +
geom_histogram(mapping = aes(x = y), binwidth = 0.5)
ggplot(table) +
geom_histogram(mapping = aes(x = y), binwidth = 0.5) +
coord_cartesian(ylim = c(0, 50))
# unusual <- table %>% # filter(y < 30 | y > 60) %>% # select(x, y) %>% # arrange(y) # unusual
ggplot(data = table, mapping = aes(x = x, y = y)) +
geom_point()
ggplot(data = table, mapping = aes(x = x, y = y)) +
geom_point(na.rm = TRUE)
ggplot(data = table, mapping = aes(x = precision)) +
geom_freqpoly(mapping = aes(colour = classifier), binwidth = 500)
ggplot(data = table, mapping = aes(x = precision)) +
geom_freqpoly(mapping = aes(colour = classifier), binwidth = 500)
ggplot(data = table, mapping = aes(x = precision, y = ..recall..)) +
geom_freqpoly(mapping = aes(colour = classifier), binwidth = 500)
ggplot(data = table, mapping = aes(x = precision, y = recall)) +
geom_boxplot()
ggplot(data = table, mapping = aes(x = classifier, y = precision)) +
geom_boxplot()
ggplot(data = table, mapping = aes(x = classifier, y = recall)) +
geom_boxplot()
ggplot(data = table) +
geom_boxplot(mapping = aes(x = reorder(classifier, precision, FUN = median), y = precision))
ggplot(data = table) +
geom_boxplot(mapping = aes(x = reorder(classifier, precision, FUN = median), y = precision)) +
coord_flip()
ggplot(data = table) +
geom_count(mapping = aes(x = classifier, y = sampling))
table %>%
count(Classifier, Sampling)## Classifier Sampling n
## 1 DecisionTree FS 1
## 2 DecisionTree Imbalanced 9
## 3 DecisionTree Oversampling 18
## 4 DecisionTree Undersampling 27
## 5 Logistic Reg FS 1
## 6 Logistic Reg Imbalanced 9
## 7 Logistic Reg Oversampling 18
## 8 Logistic Reg Undersampling 27
## 9 Naive Bayes FS 1
## 10 Naive Bayes Imbalanced 9
## 11 Naive Bayes Oversampling 18
## 12 Naive Bayes Undersampling 27
## 13 Random Forest Imbalanced 1
## 14 Random Forest Oversampling 2
## 15 Random Forest Undersampling 2
## 16 Random Forest Unsersampling 1
## 17 XGBoost Imbalanced 9
## 18 XGBoost Oversampling 18
## 19 XGBoost Undersampling 27table %>%
count(Classifier, Sampling) %>%
ggplot(mapping = aes(x = Classifier, y = Sampling)) +
geom_tile(mapping = aes(fill = n))
ggplot(data = table) +
geom_point(mapping = aes(x = Precision, y = Recall), alpha = 50 / 100)
ggplot(data = smaller) +
geom_bin2d(mapping = aes(x = Precision, y = Recall))
# install.packages("hexbin")
library(hexbin)ggplot(data = smaller) +
geom_hex(mapping = aes(x = Precision, y = Recall))
ggplot(data = smaller, mapping = aes(x = Precision, y = Recall)) +
geom_boxplot(mapping = aes(group = cut_width(Precision, 0.1)))
ggplot(data = smaller, mapping = aes(x = Precision, y = Recall)) +
geom_boxplot(mapping = aes(group = cut_number(Precision, 20)))
ggplot(data = table) +
geom_point(mapping = aes(x = x, y = y)) +
coord_cartesian(xlim = c(0, 90), ylim = c(0, 90))
ggplot(data = table) +
geom_point(mapping = aes(x = Precision, y = Recall))
library(modelr)mod <- lm(log(Recall) ~ log(Precision), data = table)
table1 <- table %>%
add_residuals(mod) %>%
mutate(resid = exp(resid))ggplot(data = table1) +
geom_point(mapping = aes(x = Precision, y = resid))
ggplot(data = table1) +
geom_boxplot(mapping = aes(x = Precision, y = resid))
ggplot(data = table, mapping = aes(x = precision)) +
geom_freqpoly(binwidth = 0.25)
ggplot(table, aes(Precision)) +
geom_freqpoly(binwidth = 0.25)
table %>%
count(Classifier, Sampling) %>%
ggplot(aes(Classifier, Sampling, fill = n)) +
geom_tile()
Chapter 9 Wrangle
table %>%
count(Precision) %>%
filter(n > 1)## Precision n
## 1 21.81 2
## 2 22.65 2
## 3 24.91 2
## 4 28.27 2
## 5 57.47 2
## 6 58.71 2
## 7 58.76 3
## 8 59.57 2
## 9 60.53 2
## 10 60.96 2
## 11 83.06 2
## 12 84.04 2
## 13 94.27 2table %>%
count(Recall) %>%
filter(n > 1)## Recall n
## 1 14.30 2
## 2 46.47 2
## 3 69.45 2
## 4 88.06 2
## 5 88.45 2
## 6 88.66 2
## 7 88.80 2
## 8 89.86 2
## 9 92.82 2
## 10 93.67 2table %>%
count(Precision, Recall) %>%
filter(n > 1)## Precision Recall n
## 1 22.65 69.45 2
## 2 28.27 88.45 2
## 3 57.47 88.06 2
## 4 58.71 89.86 2
## 5 58.76 88.66 2
## 6 60.53 92.82 2
## 7 60.96 93.67 2
## 8 83.06 46.47 2
## 9 84.04 14.30 2Chapter 23 Model Basics
x <- precision
y <- recall
# models <- tibble(
# a1 = c(precision),
# a2 = c(recall)
# )models <- tibble(
a1 = precision,
a2 = recall
)
models## # A tibble: 225 × 2
## a1 a2
## <dbl> <dbl>
## 1 74.5 16.7
## 2 72.8 21.2
## 3 12.7 75.0
## 4 59 22.8
## 5 22.6 69.4
## 6 76.7 35.4
## 7 76.4 34.8
## 8 26.0 82.9
## 9 77.5 31
## 10 22.6 69.4
## # … with 215 more rowsggplot(table, aes(x=precision, y=recall)) +
geom_abline(aes(intercept = a1, slope = a2), data = models, alpha = 1/4) +
geom_point()
model1 <- function(a, data) {
a[1] + data$x * a[2]
}
# model1(c(7, 1.5), table)
#> [1] 8.5 8.5 8.5 10.0 10.0 10.0 11.5 11.5 11.5 13.0 13.0 13.0 14.5 14.5 14.5
#> [16] 16.0 16.0 16.0 17.5 17.5 17.5 19.0 19.0 19.0 20.5 20.5 20.5 22.0 22.0 22.0
measure_distance <- function(mod, data) {
diff <- data$y - model1(mod, data)
sqrt(mean(diff ^ 2))
}
table_dist <- function(a1, a2) {
measure_distance(c(a1, a2), table)
}
# test2 <- c(precision[1],recall[2])
# test2
# measure_distance(c(7.5, 1), table)
models <- models %>%
mutate(dist = purrr::map2_dbl(a1, a2, table_dist))
models## # A tibble: 225 × 3
## a1 a2 dist
## <dbl> <dbl> <dbl>
## 1 74.5 16.7 NaN
## 2 72.8 21.2 NaN
## 3 12.7 75.0 NaN
## 4 59 22.8 NaN
## 5 22.6 69.4 NaN
## 6 76.7 35.4 NaN
## 7 76.4 34.8 NaN
## 8 26.0 82.9 NaN
## 9 77.5 31 NaN
## 10 22.6 69.4 NaN
## # … with 215 more rows#> # A tibble: 250 x 3 #> a1 a2 dist #>
ggplot(table, aes(x = precision, y = recall)) +
geom_point(size = 2, colour = "grey30") +
geom_abline(
aes(intercept = a1, slope = a2, colour = -dist),
data = filter(models, rank(dist) <= 10)
)
ggplot(models, aes(a1, a2)) +
geom_point(data = filter(models, rank(dist) <= 10), size = 4, colour = "red") +
geom_point(aes(colour = -dist))
grid <- expand.grid(
a1 = seq(-5, 20, length=25, along.with=precision),
a2 = seq(1, 3, length = 25, along.with=recall)
) %>%
mutate(dist = purrr::map2_dbl(a1, a2, table_dist))grid %>%
ggplot(aes(a1, a2)) +
geom_point(data = filter(grid, rank(dist) <= 10), size = 4, colour = "red") +
geom_point(aes(colour = -dist)) 
ggplot(table, aes(x=precision, y=recall)) +
geom_point(size = 2, colour = "grey30") +
geom_abline(
aes(intercept = a1, slope = a2, colour = -dist),
data = filter(grid, rank(dist) <= 10)
)
# best <- optim(c(0, 0), measure_distance, data = table) # best$par
ggplot(table, aes(x, y)) +
geom_point(size = 2, colour = “grey30”) +
geom_abline(intercept = best\(par[1], slope = best\)par[2])
table_mod <- lm(y ~ x, data = table)
coef(table_mod)## (Intercept) x
## 95.3192201 -0.6232934tablea <- tibble(
x = rep(1:10, each = 3),
y = x * 1.5 + 6 + rt(length(x), df = 2)
)
measure_distance <- function(mod, data) {
diff <- data$y - model1(mod, data)
mean(abs(diff))
}
model1 <- function(a, data) {
a[1] + data$x * a[2] + a[3]
}
grid <- table %>%
data_grid(x)
grid## # A tibble: 211 × 1
## x
## <dbl>
## 1 0.74
## 2 8.31
## 3 8.44
## 4 9.32
## 5 9.95
## 6 10.5
## 7 10.8
## 8 12.1
## 9 12.7
## 10 12.8
## # … with 201 more rows#> # A tibble: 10 x 1
#> x
#> <int>
#> 1 1
#> 2 2
#> 3 3
#> 4 4
#> 5 5
#> 6 6
#> # . with 4 more rows
grid <- grid %>%
add_predictions(table_mod)
grid## # A tibble: 211 × 2
## x pred
## <dbl> <dbl>
## 1 0.74 94.9
## 2 8.31 90.1
## 3 8.44 90.1
## 4 9.32 89.5
## 5 9.95 89.1
## 6 10.5 88.7
## 7 10.8 88.6
## 8 12.1 87.8
## 9 12.7 87.4
## 10 12.8 87.4
## # … with 201 more rows#> # A tibble: 10 x 2
#> x pred
#> <int> <dbl>
#> 1 1 6.27
#> 2 2 8.32
#> 3 3 10.4
#> 4 4 12.4
#> 5 5 14.5
#> 6 6 16.5
#> # . with 4 more rows
ggplot(table, aes(x)) +
geom_point(aes(y = y)) +
geom_line(aes(y = pred), data = grid, colour = "red", size = 1)
table <- table %>%
add_residuals(table_mod)
table## Year Sampling Technique Classifier Precision Recall
## 1 2010 Imbalanced N/A Naive Bayes 74.49 16.70
## 2 2010 Imbalanced N/A Logistic Reg 72.82 21.18
## 3 2010 Imbalanced N/A XGBoost 12.66 75.05
## 4 2010 Imbalanced N/A DecisionTree 59.00 22.77
## 5 2010 Imbalanced N/A Random Forest 22.65 69.45
## 6 2010 Undersampling NearMiss Naive Bayes 76.67 35.38
## 7 2010 Undersampling NearMiss Logistic Reg 76.37 34.83
## 8 2010 Undersampling NearMiss XGBoost 26.04 82.92
## 9 2010 Undersampling NearMiss DecisionTree 77.47 31.00
## 10 2010 Undersampling NearMiss Random Forest 22.65 69.45
## 11 2010 Oversampling SMOTE Naive Bayes 57.98 88.41
## 12 2010 Oversampling SMOTE Logistic Reg 58.15 88.40
## 13 2010 Oversampling SMOTE XGBoost 94.16 54.34
## 14 2010 Oversampling SMOTE DecisionTree 56.43 87.71
## 15 2010 Oversampling SMOTE Random Forest 58.17 90.11
## 16 2010 Oversampling ROS Naive Bayes 57.95 89.05
## 17 2010 Oversampling ROS Logistic Reg 57.91 88.23
## 18 2010 Oversampling ROS XGBoost 94.52 54.08
## 19 2010 Oversampling ROS DecisionTree 56.46 88.55
## 20 2010 Oversampling ROS Random Forest 57.68 89.70
## 21 2010 Undersampling RUS Naive Bayes 67.54 35.25
## 22 2010 Undersampling RUS Logistic Reg 57.29 86.59
## 23 2010 Undersampling RUS XGBoost 93.31 53.88
## 24 2010 Undersampling RUS DecisionTree 54.84 85.36
## 25 2010 Unsersampling RUS Random Forest 58.32 86.82
## 26 2010 Undersampling Tomelinks Naive Bayes 58.00 88.92
## 27 2010 Undersampling Tomelinks Logistic Reg 73.39 35.93
## 28 2010 Undersampling Tomelinks XGBoost 94.27 54.13
## 29 2010 Undersampling Tomelinks DecisionTree 56.52 88.80
## 30 2010 Undersampling Tomelinks Random Forest 57.96 89.30
## 31 2012 Imbalanced N/A Naive Bayes 82.95 12.68
## 32 2012 Imbalanced N/A Logistic Reg 80.00 18.73
## 33 2012 Imbalanced N/A XGBoost 13.06 80.78
## 34 2012 Imbalanced N/A DecisionTree 62.30 23.00
## 35 2012 Undersampling NearMiss Naive Bayes 91.79 40.57
## 36 2012 Undersampling NearMiss Logistic Reg 91.84 39.06
## 37 2012 Undersampling NearMiss XGBoost 30.11 97.70
## 38 2012 Undersampling NearMiss DecisionTree 95.96 36.43
## 39 2012 Oversampling SMOTE Naive Bayes 58.71 89.86
## 40 2012 Oversampling SMOTE Logistic Reg 58.76 88.66
## 41 2012 Oversampling SMOTE XGBoost 21.81 84.12
## 42 2012 Oversampling SMOTE DecisionTree 57.47 88.06
## 43 2012 Oversampling ROS Naive Bayes 76.52 42.70
## 44 2012 Oversampling ROS Logistic Reg 77.85 40.72
## 45 2012 Oversampling ROS XGBoost 22.85 79.33
## 46 2012 Oversampling ROS DecisionTree 79.77 39.89
## 47 2012 Undersampling RUS Naive Bayes 72.59 37.17
## 48 2012 Undersampling RUS Logistic Reg 73.62 34.90
## 49 2012 Undersampling RUS XGBoost 22.10 75.75
## 50 2012 Undersampling RUS DecisionTree 69.89 28.39
## 51 2012 Undersampling Tomelinks Naive Bayes 59.57 89.88
## 52 2012 Undersampling Tomelinks Logistic Reg 77.36 41.44
## 53 2012 Undersampling Tomelinks XGBoost 23.72 79.00
## 54 2012 Undersampling Tomelinks DecisionTree 78.82 40.96
## 55 2013 Imbalanced N/A Naive Bayes 74.70 13.27
## 56 2013 Imbalanced N/A Logistic Reg 73.53 19.52
## 57 2013 Imbalanced N/A XGBoost 12.76 76.85
## 58 2013 Imbalanced N/A DecisionTree 56.04 22.23
## 59 2013 Undersampling NearMiss Naive Bayes 90.40 41.79
## 60 2013 Undersampling NearMiss Logistic Reg 89.11 40.89
## 61 2013 Undersampling NearMiss XGBoost 30.99 97.43
## 62 2013 Undersampling NearMiss DecisionTree 94.91 37.03
## 63 2013 Oversampling SMOTE Naive Bayes 58.71 89.86
## 64 2013 Oversampling SMOTE Logistic Reg 58.76 88.66
## 65 2013 Oversampling SMOTE XGBoost 21.81 84.13
## 66 2013 Oversampling SMOTE DecisionTree 57.47 88.06
## 67 2013 Oversampling ROS Naive Bayes 75.12 43.64
## 68 2013 Oversampling ROS Logistic Reg 75.63 42.43
## 69 2013 Oversampling ROS XGBoost 24.54 76.83
## 70 2013 Oversampling ROS DecisionTree 74.68 40.05
## 71 2013 Undersampling RUS Naive Bayes 71.84 39.03
## 72 2013 Undersampling RUS Logistic Reg 71.37 37.24
## 73 2013 Undersampling RUS XGBoost 23.37 73.38
## 74 2013 Undersampling RUS DecisionTree 66.15 30.01
## 75 2013 Undersampling Tomelinks Naive Bayes 59.20 89.18
## 76 2013 Undersampling Tomelinks Logistic Reg 75.19 42.75
## 77 2013 Undersampling Tomelinks XGBoost 24.91 76.92
## 78 2013 Undersampling Tomelinks DecisionTree 74.29 40.53
## 79 2014 Imbalanced N/A Naive Bayes 84.51 9.13
## 80 2014 Imbalanced N/A Logistic Reg 84.04 14.30
## 81 2014 Imbalanced N/A XGBoost 10.54 83.97
## 82 2014 Imbalanced N/A DecisionTree 62.37 21.66
## 83 2014 Undersampling NearMiss Naive Bayes 49.33 88.80
## 84 2014 Undersampling NearMiss Logistic Reg 84.04 14.30
## 85 2014 Undersampling NearMiss XGBoost 29.21 97.28
## 86 2014 Undersampling NearMiss DecisionTree 95.63 35.76
## 87 2014 Oversampling SMOTE Naive Bayes 59.31 91.58
## 88 2014 Oversampling SMOTE Logistic Reg 59.65 91.14
## 89 2014 Oversampling SMOTE XGBoost 95.00 54.98
## 90 2014 Oversampling SMOTE DecisionTree 58.76 91.31
## 91 2014 Oversampling ROS Naive Bayes 75.44 44.17
## 92 2014 Oversampling ROS Logistic Reg 80.77 41.96
## 93 2014 Oversampling ROS XGBoost 23.00 82.52
## 94 2014 Oversampling ROS DecisionTree 82.50 43.45
## 95 2014 Undersampling RUS Naive Bayes 57.43 90.28
## 96 2014 Undersampling RUS Logistic Reg 72.78 35.83
## 97 2014 Undersampling RUS XGBoost 94.27 53.47
## 98 2014 Undersampling RUS DecisionTree 54.68 86.53
## 99 2014 Undersampling Tomelinks Naive Bayes 75.27 43.66
## 100 2014 Undersampling Tomelinks Logistic Reg 80.05 42.08
## 101 2014 Undersampling Tomelinks XGBoost 23.23 82.18
## 102 2014 Undersampling Tomelinks DecisionTree 59.33 93.08
## 103 2014 FS Standard Scalar DecisionTree 61.90 21.62
## 104 2014 FS Extra Tree Naive Bayes 66.09 24.14
## 105 2014 FS SS & SKB Logistic Reg 86.79 1.42
## 106 2015 Imbalanced N/A Naive Bayes 82.11 6.44
## 107 2015 Imbalanced N/A Logistic Reg 78.39 12.22
## 108 2015 Imbalanced N/A XGBoost 9.32 80.94
## 109 2015 Imbalanced N/A DecisionTree 58.11 21.06
## 110 2015 Undersampling NearMiss Naive Bayes 48.34 88.34
## 111 2015 Undersampling NearMiss Logistic Reg 88.13 40.58
## 112 2015 Undersampling NearMiss XGBoost 31.02 96.18
## 113 2015 Undersampling NearMiss DecisionTree 93.00 36.82
## 114 2015 Oversampling SMOTE Naive Bayes 59.16 90.34
## 115 2015 Oversampling SMOTE Logistic Reg 59.19 90.36
## 116 2015 Oversampling SMOTE XGBoost 24.50 83.16
## 117 2015 Oversampling SMOTE DecisionTree 79.51 42.27
## 118 2015 Oversampling ROS Naive Bayes 74.87 47.24
## 119 2015 Oversampling ROS Logistic Reg 79.57 45.13
## 120 2015 Oversampling ROS XGBoost 25.46 79.86
## 121 2015 Oversampling ROS DecisionTree 80.83 46.89
## 122 2015 Undersampling RUS Naive Bayes 73.66 39.95
## 123 2015 Undersampling RUS Logistic Reg 73.87 38.19
## 124 2015 Undersampling RUS XGBoost 23.48 74.19
## 125 2015 Undersampling RUS DecisionTree 70.75 32.26
## 126 2015 Undersampling Tomelinks Naive Bayes 78.06 46.37
## 127 2015 Undersampling Tomelinks Logistic Reg 78.63 45.33
## 128 2015 Undersampling Tomelinks XGBoost 26.37 78.69
## 129 2015 Undersampling Tomelinks DecisionTree 80.56 46.69
## 130 2016 Imbalanced N/A Naive Bayes 8.44 0.25
## 131 2016 Imbalanced N/A Logistic Reg 30.32 1.33
## 132 2016 Imbalanced N/A XGBoost 0.74 18.29
## 133 2016 Imbalanced N/A DecisionTree 12.13 3.89
## 134 2016 Undersampling NearMiss Naive Bayes 48.45 87.85
## 135 2016 Undersampling NearMiss Logistic Reg 87.21 36.76
## 136 2016 Undersampling NearMiss XGBoost 28.26 93.04
## 137 2016 Undersampling NearMiss DecisionTree 90.49 32.91
## 138 2016 Oversampling SMOTE Naive Bayes 56.65 87.29
## 139 2016 Oversampling SMOTE Logistic Reg 57.18 87.52
## 140 2016 Oversampling SMOTE XGBoost 93.65 54.06
## 141 2016 Oversampling SMOTE DecisionTree 55.60 85.72
## 142 2016 Oversampling ROS Naive Bayes 57.52 91.43
## 143 2016 Oversampling ROS Logistic Reg 70.99 41.93
## 144 2016 Oversampling ROS XGBoost 24.10 70.74
## 145 2016 Oversampling ROS DecisionTree 55.83 88.20
## 146 2016 Undersampling RUS Naive Bayes 62.03 38.16
## 147 2016 Undersampling RUS Logistic Reg 63.96 31.90
## 148 2016 Undersampling RUS XGBoost 18.65 64.27
## 149 2016 Undersampling RUS DecisionTree 51.68 80.64
## 150 2016 Undersampling Tomelinks Naive Bayes 68.65 43.90
## 151 2016 Undersampling Tomelinks Logistic Reg 71.44 42.05
## 152 2016 Undersampling Tomelinks XGBoost 23.60 71.46
## 153 2016 Undersampling Tomelinks DecisionTree 55.92 89.48
## 154 2017 Imbalanced N/A Naive Bayes 82.28 7.92
## 155 2017 Imbalanced N/A Logistic Reg 79.39 15.98
## 156 2017 Imbalanced N/A XGBoost 10.83 79.20
## 157 2017 Imbalanced N/A DecisionTree 58.87 22.11
## 158 2017 Undersampling NearMiss Naive Bayes 48.98 89.36
## 159 2017 Undersampling NearMiss Logistic Reg 93.06 39.56
## 160 2017 Undersampling NearMiss XGBoost 97.36 30.88
## 161 2017 Undersampling NearMiss DecisionTree 94.23 37.48
## 162 2017 Oversampling SMOTE Naive Bayes 57.74 91.47
## 163 2017 Oversampling SMOTE Logistic Reg 57.82 90.42
## 164 2017 Oversampling SMOTE XGBoost 22.55 83.77
## 165 2017 Oversampling SMOTE DecisionTree 77.72 40.32
## 166 2017 Oversampling ROS Naive Bayes 76.68 44.67
## 167 2017 Oversampling ROS Logistic Reg 77.66 43.73
## 168 2017 Oversampling ROS XGBoost 25.33 78.13
## 169 2017 Oversampling ROS DecisionTree 78.78 44.65
## 170 2017 Undersampling RUS Naive Bayes 70.85 39.82
## 171 2017 Undersampling RUS Logistic Reg 72.51 36.51
## 172 2017 Undersampling RUS XGBoost 22.63 74.59
## 173 2017 Undersampling RUS DecisionTree 68.33 30.97
## 174 2017 Undersampling Tomelinks Naive Bayes 58.69 90.82
## 175 2017 Undersampling Tomelinks Logistic Reg 77.32 43.37
## 176 2017 Undersampling Tomelinks XGBoost 24.40 78.95
## 177 2017 Undersampling Tomelinks DecisionTree 78.95 44.18
## 178 2018 Imbalanced N/A Naive Bayes 79.06 4.85
## 179 2018 Imbalanced N/A Logistic Reg 91.19 39.00
## 180 2018 Imbalanced N/A XGBoost 8.31 83.80
## 181 2018 Imbalanced N/A DecisionTree 56.80 11.33
## 182 2018 Undersampling NearMiss Naive Bayes 91.44 41.73
## 183 2018 Undersampling NearMiss Logistic Reg 85.23 10.28
## 184 2018 Undersampling NearMiss XGBoost 30.41 97.00
## 185 2018 Undersampling NearMiss DecisionTree 94.46 35.84
## 186 2018 Oversampling SMOTE Naive Bayes 59.13 91.79
## 187 2018 Oversampling SMOTE Logistic Reg 59.94 90.74
## 188 2018 Oversampling SMOTE XGBoost 95.68 55.37
## 189 2018 Oversampling SMOTE DecisionTree 59.07 92.05
## 190 2018 Oversampling ROS Naive Bayes 59.51 93.60
## 191 2018 Oversampling ROS Logistic Reg 80.01 43.26
## 192 2018 Oversampling ROS XGBoost 24.91 82.35
## 193 2018 Oversampling ROS DecisionTree 59.57 94.69
## 194 2018 Undersampling RUS Naive Bayes 57.13 90.33
## 195 2018 Undersampling RUS Logistic Reg 72.71 35.90
## 196 2018 Undersampling RUS XGBoost 22.68 74.60
## 197 2018 Undersampling RUS DecisionTree 68.43 29.72
## 198 2018 Undersampling Tomelinks Naive Bayes 76.05 44.14
## 199 2018 Undersampling Tomelinks Logistic Reg 79.88 43.31
## 200 2018 Undersampling Tomelinks XGBoost 23.90 82.25
## 201 2018 Undersampling Tomelinks DecisionTree 82.42 45.59
## 202 2019 Imbalanced N/A Naive Bayes 80.14 3.79
## 203 2019 Imbalanced N/A Logistic Reg 85.94 10.82
## 204 2019 Imbalanced N/A XGBoost 9.95 86.31
## 205 2019 Imbalanced N/A DecisionTree 59.93 20.26
## 206 2019 Undersampling NearMiss Naive Bayes 91.34 44.56
## 207 2019 Undersampling NearMiss Logistic Reg 92.64 40.48
## 208 2019 Undersampling NearMiss XGBoost 32.26 96.06
## 209 2019 Undersampling NearMiss DecisionTree 94.48 37.98
## 210 2019 Oversampling SMOTE Naive Bayes 60.53 92.82
## 211 2019 Oversampling SMOTE Logistic Reg 60.96 93.67
## 212 2019 Oversampling SMOTE XGBoost 28.27 88.45
## 213 2019 Oversampling SMOTE DecisionTree 83.06 46.47
## 214 2019 Oversampling ROS Naive Bayes 60.53 92.82
## 215 2019 Oversampling ROS Logistic Reg 60.96 93.67
## 216 2019 Oversampling ROS XGBoost 28.27 88.45
## 217 2019 Oversampling ROS DecisionTree 83.06 46.47
## 218 2019 Undersampling RUS Naive Bayes 71.22 45.77
## 219 2019 Undersampling RUS Logistic Reg 75.98 30.80
## 220 2019 Undersampling RUS XGBoost 27.22 76.95
## 221 2019 Undersampling RUS DecisionTree 72.62 32.97
## 222 2019 Undersampling Tomelinks Naive Bayes 81.20 48.30
## 223 2019 Undersampling Tomelinks Logistic Reg 82.89 48.10
## 224 2019 Undersampling Tomelinks XGBoost 29.04 84.70
## 225 2019 Undersampling Tomelinks DecisionTree 85.60 49.16
## resid
## 1 -32.19009215
## 2 -28.75099218
## 3 -12.37832520
## 4 -35.77490745
## 5 -11.75162379
## 6 -12.15131246
## 7 -12.88830049
## 8 3.83134096
## 9 -16.03267771
## 10 -11.75162379
## 11 29.22933325
## 12 29.32529313
## 13 17.71008971
## 14 27.56322843
## 15 31.04775900
## 16 29.85063445
## 17 29.00570271
## 18 17.67447534
## 19 28.42192723
## 20 30.33234522
## 21 -17.97198152
## 22 26.97926078
## 23 16.72029029
## 24 24.22219187
## 25 27.85125302
## 26 29.75179912
## 27 -13.64571493
## 28 17.56865198
## 29 28.70932484
## 30 30.10686738
## 31 -30.93702969
## 32 -26.72574533
## 33 -6.39900782
## 34 -33.48803911
## 35 2.46288427
## 36 0.98404894
## 37 21.14814523
## 38 0.92201789
## 39 31.13433746
## 40 29.96550213
## 41 2.39480973
## 42 28.56145360
## 43 -4.92480648
## 44 -6.07582621
## 45 -1.74696510
## 46 -5.70910282
## 47 -12.90434967
## 48 -14.53235744
## 49 -5.79443518
## 50 -23.36724195
## 51 31.69036981
## 52 -5.66123999
## 53 -1.53469981
## 54 -5.23123158
## 55 -35.48920053
## 56 -29.96845385
## 57 -10.51599585
## 58 -38.15985601
## 59 2.81650639
## 60 1.11245786
## 61 21.42664346
## 62 0.86755978
## 63 31.13433746
## 64 29.96550213
## 65 2.40480973
## 66 28.56145360
## 67 -4.85741728
## 68 -5.74953763
## 69 -3.19359920
## 70 -8.72166640
## 71 -11.51181975
## 72 -13.59476766
## 73 -7.37285251
## 74 -24.07835939
## 75 30.75975124
## 76 -5.70378674
## 77 -2.87298062
## 78 -8.48475084
## 79 -33.51469194
## 80 -28.63763985
## 81 -4.77970728
## 82 -34.78440857
## 83 24.22784504
## 84 -28.63763985
## 85 20.16718114
## 86 0.04633105
## 87 33.22831352
## 88 33.00023329
## 89 18.87365619
## 90 32.61550213
## 91 -4.12796339
## 92 -3.01580938
## 93 1.53652892
## 94 -0.44751174
## 95 30.75652186
## 96 -14.12592392
## 97 16.90865198
## 98 25.29246492
## 99 -4.74392327
## 100 -3.34458065
## 101 1.33988641
## 102 34.74077939
## 103 -35.11735649
## 104 -29.98575700
## 105 -39.80358291
## 106 -37.70059618
## 107 -34.23924775
## 108 -8.57012527
## 109 -38.03963860
## 110 23.15078454
## 111 0.19163030
## 112 20.19534226
## 113 -0.53293068
## 114 31.89481950
## 115 31.93351831
## 116 3.11146907
## 117 -3.49115911
## 118 -1.41324064
## 119 -0.59376150
## 120 0.40983076
## 121 1.95158823
## 122 -9.45742570
## 123 -11.08653408
## 124 -6.49429024
## 125 -18.96120959
## 126 -0.29493459
## 127 -0.97965733
## 128 -0.19297221
## 129 1.58329900
## 130 -89.80862349
## 131 -75.09096314
## 132 -76.56798293
## 133 -83.86867072
## 134 22.72934682
## 135 -4.20179966
## 136 15.33505238
## 137 -6.00739720
## 138 27.28035298
## 139 27.84069850
## 140 17.11221005
## 141 25.05589488
## 142 31.96261827
## 143 -9.14161917
## 144 -9.55784831
## 145 27.67925237
## 146 -18.49632834
## 147 -23.55337201
## 148 -19.42479752
## 149 17.53258461
## 150 -8.63012581
## 151 -8.74113712
## 152 -9.14949502
## 153 29.01534878
## 154 -36.11463629
## 155 -29.85595432
## 156 -9.36895218
## 157 -36.51593559
## 158 24.56969234
## 159 2.24446693
## 160 -3.75537130
## 161 0.89372025
## 162 32.13974283
## 163 31.13960630
## 164 2.50604687
## 165 -6.55685436
## 166 -2.85507953
## 167 -3.18425196
## 168 -1.40119738
## 169 -1.56616332
## 170 -11.33888025
## 171 -13.61421315
## 172 -6.62408966
## 173 -21.75957970
## 174 32.08187159
## 175 -3.75617173
## 176 -1.16086028
## 177 -1.93020343
## 178 -41.19164115
## 179 0.51890821
## 180 -6.33965164
## 181 -48.58615300
## 182 3.40473156
## 183 -31.91592066
## 184 20.63513326
## 185 -0.60292226
## 186 33.32612070
## 187 32.78098838
## 188 19.68749573
## 189 33.54872309
## 190 35.37297220
## 191 -2.18951239
## 192 2.55701938
## 193 36.50036981
## 194 30.61953383
## 195 -14.09955446
## 196 -6.58292498
## 197 -22.94725036
## 198 -3.77775439
## 199 -2.22054054
## 200 1.82749301
## 201 1.64262479
## 202 -41.57848424
## 203 -30.93338232
## 204 -2.80745040
## 205 -37.70524455
## 206 6.17240222
## 207 2.90268369
## 208 20.84822612
## 209 1.54954360
## 210 35.22873151
## 211 36.34674768
## 212 10.75128531
## 213 2.92153258
## 214 35.22873151
## 215 36.34674768
## 216 10.75128531
## 217 2.92153258
## 218 -5.15826168
## 219 -17.16138493
## 220 -1.40317279
## 221 -17.08565087
## 222 3.59220680
## 223 4.44557270
## 224 7.48122126
## 225 7.19469791#> # A tibble: 30 x 3
#> x y resid
#> <int> <dbl> <dbl>
#> 1 1 4.20 -2.07
#> 2 1 7.51 1.24
#> 3 1 2.13 -4.15
#> 4 2 8.99 0.665
#> 5 2 10.2 1.92
#> 6 2 11.3 2.97
#> # . with 24 more rows
ggplot(table, aes(resid)) +
geom_freqpoly(binwidth = 0.5)
ggplot(table, aes(x, resid)) +
geom_ref_line(h = 0) +
geom_point() 
ggplot(table) +
geom_point(aes(x, y))
mod2 <- lm(y ~ x, data = table)
grid <- table %>%
data_grid(x) %>%
add_predictions(mod2)
grid## # A tibble: 211 × 2
## x pred
## <dbl> <dbl>
## 1 0.74 94.9
## 2 8.31 90.1
## 3 8.44 90.1
## 4 9.32 89.5
## 5 9.95 89.1
## 6 10.5 88.7
## 7 10.8 88.6
## 8 12.1 87.8
## 9 12.7 87.4
## 10 12.8 87.4
## # … with 201 more rows#> # A tibble: 4 x 2
#> x pred
#> <chr> <dbl>
#> 1 a 1.15
#> 2 b 8.12
#> 3 c 6.13
#> 4 d 1.91
ggplot(table, aes(x)) +
geom_point(aes(y = y)) +
geom_point(data = grid, aes(y = pred), colour = "red", size = 4)
# tibble(x = "e") %>%
# add_predictions(mod2)
ggplot(table, aes(x, y)) +
geom_point(aes(colour = classifier))
#> Error in model.frame.default(Terms, newdata, na.action = na.action, xlev = object$xlevels): factor x has new level emod1 <- lm(y ~ x + classifier, data = table)
mod2 <- lm(y ~ x * classifier, data = table)
grid <- table %>%
data_grid(x, classifier) %>%
gather_predictions(mod1, mod2)
grid## # A tibble: 2,110 × 4
## model x classifier pred
## <chr> <dbl> <chr> <dbl>
## 1 mod1 0.74 DecisionTree 85.1
## 2 mod1 0.74 Logistic Reg 83.9
## 3 mod1 0.74 Naive Bayes 91.5
## 4 mod1 0.74 Random Forest 106.
## 5 mod1 0.74 XGBoost 93.6
## 6 mod1 8.31 DecisionTree 81.1
## 7 mod1 8.31 Logistic Reg 79.9
## 8 mod1 8.31 Naive Bayes 87.5
## 9 mod1 8.31 Random Forest 102.
## 10 mod1 8.31 XGBoost 89.6
## # … with 2,100 more rows# Problematic Code
foo = tibble(x=precision, y=recall, classifier=classifier, sampling=sampling, technique=technique, year=year)
ggplot(foo, aes(x, y, colour = classifier)) +
geom_point() +
geom_line(data = grid, aes(y = pred)) +
facet_wrap(~ model)
Leftover Code
#> # A tibble: 80 x 4 #> model x1 x2 pred #>
# Leftover Code
#> # A tibble: 80 x 4
#> model x1 x2 pred
#> <chr> <int> <fct> <dbl>
#> 1 mod1 1 a 1.67
#> 2 mod1 1 b 4.56
#> 3 mod1 1 c 6.48
#> 4 mod1 1 d 4.03
#> 5 mod1 2 a 1.48
#> 6 mod1 2 b 4.37
#> # . with 74 more rows
ggplot(table, aes(x=precision, y=recall)) +
geom_point(size = 2, colour = "grey30")
# ggplot(foo, aes(x, y)) + geom_boxplot()
# ggplot(classifier, sampling, aes(x, y)) + geom_boxplot()
# # best <- optim(c(0, 0), measure_distance, data = table)
# best$par
# #> [1] 4.222248 2.051204
#
# data_dist <- function(x, y) {
# measure_distance(c(x, y), table)
# }
#
# models <- foo %>%
# mutate(dist = purrr::foo(x, y, data_dist))
# models
# ggplot(table, aes(x=precision, y=recall)) +
# geom_point(size = 2, colour = "grey30") +
# geom_abline(intercept = best$par[1], slope = best$par[2])
# ggplot(classifier, aes(color, price)) + geom_boxplot()
# ggplot(classifier, diamonds, aes(clarity, price)) + geom_boxplot()
#
ggplot(foo, aes(x, y)) +
geom_point(size = 2, colour = "grey30") +
geom_abline(
aes(intercept = a1, slope = a2, colour = -dist),
data = filter(models, rank(dist) <= 0)
)
# grid <- expand.grid(
# a1 = seq(-5, 20, length = 25),
# a2 = seq(1, 3, length = 25)
# ) %>%
# mutate(dist = purrr::map2_dbl(x, y, data_dist))
#
# grid %>%
# ggplot(aes(x, y)) +
# geom_point(data = filter(grid, rank(dist) <= 10), size = 4, colour = "red") +
# geom_point(aes(colour = -dist))
#
foo <- foo %>%
gather_residuals(mod1, mod2)
ggplot(foo, aes(x, y, colour = classifier)) +
geom_point() +
facet_grid(model ~ classifier)
mod1 <- lm(y ~ x + classifier, data = table)
mod2 <- lm(y ~ x * classifier, data = table)mod1 <- lm(y ~ x + classifier, data = table)
mod2 <- lm(y ~ x * classifier, data = table)
# Problematic Code
# grid <- foo %>%
# data_grid(
# x = seq_range(x, 5),
# y = seq_range(y, 5)
# ) %>%
# gather_predictions(mod1, mod2)
# grid
ggplot(foo, aes(x, y)) +
geom_tile(aes(y = classifier)) +
facet_wrap(~ model)
ggplot(foo, aes(x, y, colour= classifier, group = classifier)) +
geom_line() +
facet_wrap(~ model)
ggplot(foo, aes(y, x, colour = classifier, group = classifier)) +
geom_line() +
facet_wrap(~ model)
ggplot(foo, aes(x, y)) +
geom_tile(aes(fill = classifier)) +
facet_wrap(~ model)
Chapter 24 Model Building
library(tidyverse)
library(modelr)
options(na.action = na.warn)
# library(nycflights13)
# library(lubridate)
# 24.2
ggplot(foo, aes(x, classifier)) + geom_boxplot()
ggplot(foo, aes(y, classifier)) + geom_boxplot()
# 24.2.1
ggplot(foo, aes(x, y)) +
geom_hex(bins = 50)
ggplot(foo, aes(x, classifier)) +
geom_hex(bins = 50)
ggplot(foo, aes(y, classifier)) +
geom_hex(bins = 50)
# Code to ignore
# foo <- foo %>%
# add_residuals(mod_foo, "lclassifier")
#
# ggplot(foo, aes(x, lclassifier)) +
# geom_hex(bins = 50)# grid <- foo2 %>%
# data_grid(x = seq_range(x), 20)) %>%
# mutate(x = log2(x)) %>%
# add_predictions(mod_foo2, "l_x") %>%
# mutate(x = 2 ^ x)
lm1 <- lm(y ~ classifier, data=foo)
lm2 <- lm(x ~ classifier, data=foo)
ggplot(foo, aes(x, y)) +
geom_hex(bins = 50) +
geom_line(data = foo, colour = "red", size = 1)
# foo <- foo %>%
# filter(x <= 1) %>%
# mutate(l_x = log2(l_x), l_y = log2(y))
# ggplot(foo, aes(l_x, l_y)) +
# geom_hex(bins = 50)mod_foo <- lm(x ~ y, data = foo)
foo <- foo %>%
add_residuals(mod_foo, "classifier")
ggplot(foo, aes(x, classifier)) + geom_boxplot()
ggplot(foo, aes(y, classifier)) + geom_boxplot()
foo = tibble(x=precision, y=recall, classifier=classifier, sampling=sampling, technique=technique, year=year)
mod_foo <- lm(y ~ x + classifier + sampling + technique, data = foo)
mod_foo <- lm(y ~ x, data = foo)
# grid <- foo2 %>%
# data_grid(x = seq_range(x, 225)) %>%
# mutate(l_x = log2(x)) %>%
# add_predictions(mod_foo2, "l_y") %>%
# mutate(l_y = log2(y))
foo <- foo %>%
add_residuals(mod_foo, "technique")
ggplot(foo, aes(x, technique)) +
geom_hex(bins = 50)
ggplot(foo, aes(y, technique)) +
geom_hex(bins = 50)
ggplot(foo, aes(x, technique)) + geom_boxplot()
ggplot(foo, aes(y, technique)) + geom_boxplot()
Chapter 24.2.2
foo = tibble(x=precision, y=recall, classifier=classifier, sampling=sampling, technique=technique, year=year)
mod_foo <- lm(y ~ x + classifier + sampling + technique, data = foo)
grid <- foo %>%
data_grid(classifier, .model = mod_foo) %>%
add_predictions(mod_foo)
grid## # A tibble: 30 × 5
## classifier x sampling technique pred
## <chr> <dbl> <chr> <chr> <dbl>
## 1 DecisionTree 62.3 Undersampling N/A 59.7
## 2 DecisionTree 62.3 Undersampling NearMiss 58.5
## 3 DecisionTree 62.3 Undersampling ROS 44.1
## 4 DecisionTree 62.3 Undersampling RUS 48.8
## 5 DecisionTree 62.3 Undersampling SMOTE 59.7
## 6 DecisionTree 62.3 Undersampling Tomelinks 59.7
## 7 Logistic Reg 62.3 Undersampling N/A 59.2
## 8 Logistic Reg 62.3 Undersampling NearMiss 58.0
## 9 Logistic Reg 62.3 Undersampling ROS 43.6
## 10 Logistic Reg 62.3 Undersampling RUS 48.3
## # … with 20 more rowsggplot(grid, aes(x, pred)) +
geom_point()
#> # A tibble: 5 x 5
#> cut lcarat color clarity pred
#> <ord> <dbl> <chr> <chr> <dbl>
#> 1 Fair -0.515 G VS2 11.2
#> 2 Good -0.515 G VS2 11.3
#> 3 Very Good -0.515 G VS2 11.4
#> 4 Premium -0.515 G VS2 11.4
#> 5 Ideal -0.515 G VS2 11.4
ggplot(grid, aes(classifier, pred)) +
geom_point()
# 24.2.2
foo <- foo %>%
add_residuals(mod_foo, "lclassifier")
ggplot(foo, aes(x, lclassifier)) +
geom_hex(bins = 50)
foo <- foo %>%
add_residuals(mod_foo)
ggplot(foo, aes(x, classifier)) +
geom_line()
Chapter 24.3
# foo2 %>%
# filter(abs(l_x) > 1) %>%
# add_predictions(mod_foo) %>%
# mutate(pred = pred) %>%
# select(l_x, pred, l_y:all_of(foo), x:y) %>%
# arrange(x)
ggplot(foo, aes(x, classifier)) +
geom_line()
ggplot(foo, aes(y, classifier)) +
geom_line()
ggplot(foo, aes(x, y)) +
geom_line()
# Chapter 24.3.1
ggplot(foo, aes(classifier, x)) +
geom_boxplot()
ggplot(foo, aes(classifier, y)) +
geom_boxplot()
mod <- lm(x ~ classifier, data = foo)
grid <- foo %>%
data_grid(classifier) %>%
add_predictions(mod, "x")
ggplot(foo, aes(classifier, x)) +
geom_boxplot() +
geom_point(data = grid, colour = "red", size = 4)
foo <- foo %>%
add_residuals(mod)
foo %>%
ggplot(aes(y, resid)) +
geom_ref_line(h = 0) +
geom_line()
ggplot(foo, aes(y, resid, colour = classifier)) +
geom_ref_line(h = 0) +
geom_line()
foo %>%
ggplot(aes(y, resid)) +
geom_ref_line(h = 0) +
geom_line(colour = "grey50") +
geom_smooth(se = FALSE, span = 0.20)
foo %>%
filter(classifier == "XGBoost") %>%
ggplot(aes(x, y)) +
geom_point() +
geom_line() 
foo %>%
filter(classifier == "DecisionTree") %>%
ggplot(aes(x, y)) +
geom_point() +
geom_line() 
foo %>%
filter(classifier == "Logistic Reg") %>%
ggplot(aes(x, y)) +
geom_point() +
geom_line() 
foo %>%
ggplot(aes(x, y, colour = classifier)) +
geom_boxplot()
<br
mod3 <- lm(x ~ classifier, data = foo)
mod4 <- lm(y ~ classifier, data = foo)
foo %>%
gather_residuals(precision = mod3, recall = mod4) %>%
ggplot(aes(x, resid, colour = model)) +
geom_line(alpha = 0.75)
grid <- foo %>%
data_grid(x, classifier) %>%
add_predictions(mod2, "x")
ggplot(foo, aes(classifier, x)) +
geom_boxplot() +
geom_point(data = grid, colour = "red") +
facet_wrap(~ classifier)
library(splines)
mod6 <- MASS::rlm(x ~ classifier, data = foo)
foo %>%
add_residuals(mod6, "resid") %>%
ggplot(aes(y, resid)) +
geom_hline(yintercept = 0, size = 2, colour = "white") +
geom_line()
foo %>%
data_grid(y, classifier) %>%
add_predictions(mod6) %>%
ggplot(aes(y, pred, colour = classifier)) +
geom_line() +
geom_point()
Chapter 25.1 Many Models
# foo = tibble(x=precision, y=recall, classifier=classifier, sampling=sampling, technique=technique, year=year)
foo %>%
ggplot(aes(year, precision, group = classifier)) +
geom_line(alpha = 1/3) +
ggtitle("Classifier - Precision")
foo %>%
ggplot(aes(year, recall, group = classifier)) +
geom_line(alpha = 1/3) +
ggtitle("Classifier - Recall")
foo %>%
ggplot(aes(year, precision, group = sampling)) +
geom_line(alpha = 1/3) +
ggtitle("Sampling - Precision")
foo %>%
ggplot(aes(year, recall, group = sampling)) +
geom_line(alpha = 1/3) +
ggtitle("Sampling - Recall")
foo %>%
ggplot(aes(year, precision, group = technique)) +
geom_line(alpha = 1/3) +
ggtitle("Technique - Precision")
foo %>%
ggplot(aes(year, recall, group = technique)) +
geom_line(alpha = 1/3) +
ggtitle("Technique - Recall")
# foo = tibble(x=precision, y=recall, classifier=classifier, sampling=sampling, technique=technique, year=year)
xg <- filter(foo, classifier == "XGBoost")
xg %>%
ggplot(aes(year, x)) +
geom_line() +
ggtitle("XGBoost Precision Full Data ")
# foo = tibble(x=precision, y=recall, classifier=classifier, sampling=sampling, technique=technique, year=year)
nb <- filter(foo, classifier == "Naive Bayes")
nb %>%
ggplot(aes(year, x)) +
geom_line() +
ggtitle("Naive Bayes Precision Full Data ")
# foo = tibble(x=precision, y=recall, classifier=classifier, sampling=sampling, technique=technique, year=year)
rf <- filter(foo, classifier == "Random Forest")
rf %>%
ggplot(aes(year, x)) +
geom_line() +
ggtitle("Random Forest Precision Full Data ")
# foo = tibble(x=precision, y=recall, classifier=classifier, sampling=sampling, technique=technique, year=year)
lr <- filter(foo, classifier == "Logistic Reg")
lr %>%
ggplot(aes(year, x)) +
geom_line() +
ggtitle("Logistic Regression Precision Full Data ")
# foo = tibble(x=precision, y=recall, classifier=classifier, sampling=sampling, technique=technique, year=year)
dt <- filter(foo, classifier == "DecisionTree")
dt %>%
ggplot(aes(year, x)) +
geom_line() +
ggtitle("Decision Tree Precision Full Data ")
# foo = tibble(x=precision, y=recall, classifier=classifier, sampling=sampling, technique=technique, year=year)
xg <- filter(foo, classifier == "XGBoost")
xg %>%
ggplot(aes(year, y)) +
geom_line() +
ggtitle("XGBoost Recall Full Data ")
# foo = tibble(x=precision, y=recall, classifier=classifier, sampling=sampling, technique=technique, year=year)
nb <- filter(foo, classifier == "Naive Bayes")
nb %>%
ggplot(aes(year, y)) +
geom_line() +
ggtitle("Naive Bayes Recall Full Data ")
# foo = tibble(x=precision, y=recall, classifier=classifier, sampling=sampling, technique=technique, year=year)
rf <- filter(foo, classifier == "Random Forest")
rf %>%
ggplot(aes(year, y)) +
geom_line() +
ggtitle("Random Forest Recall Full Data ")
# foo = tibble(x=precision, y=recall, classifier=classifier, sampling=sampling, technique=technique, year=year)
lr <- filter(foo, classifier == "Logistic Reg")
lr %>%
ggplot(aes(year, y)) +
geom_line() +
ggtitle("Logistic Regression Recall Full Data ")
# foo = tibble(x=precision, y=recall, classifier=classifier, sampling=sampling, technique=technique, year=year)
dt <- filter(foo, classifier == "DecisionTree")
dt %>%
ggplot(aes(year, y)) +
geom_line() +
ggtitle("Decision Tree Recall Full Data ")
xg_mod <- lm(x ~ y, classifier == "XGBoost", data = foo)
xg %>%
add_predictions(xg_mod) %>%
ggplot(aes(year, pred)) +
geom_line() +
ggtitle("XGBoost Linear trend + ")
lr_mod <- lm(x ~ y, classifier == "Logistic Reg", data = foo)
lr %>%
add_predictions(lr_mod) %>%
ggplot(aes(year, pred)) +
geom_line() +
ggtitle("Logistic Regression Linear trend + ")
rf_mod <- lm(x ~ y, classifier == "Random Forest", data = foo)
rf %>%
add_predictions(rf_mod) %>%
ggplot(aes(year, pred)) +
geom_line() +
ggtitle("Random Forest Linear trend + ")
nb_mod <- lm(x ~ y, classifier == "Naive Bayes", data = foo)
nb %>%
add_predictions(nb_mod) %>%
ggplot(aes(year, pred)) +
geom_line() +
ggtitle("Naive Bayes Linear trend + ")
dt_mod <- lm(x ~ y, classifier == "DecisionTree", data = foo)
dt %>%
add_predictions(dt_mod) %>%
ggplot(aes(year, pred)) +
geom_line() +
ggtitle("Decision Tree Linear trend + ")
xg %>%
add_residuals(xg_mod) %>%
ggplot(aes(year, resid)) +
geom_hline(yintercept = 0, colour = "white", size = 3) +
geom_line() +
ggtitle("XGBoost Remaining pattern")
lr %>%
add_residuals(lr_mod) %>%
ggplot(aes(year, resid)) +
geom_hline(yintercept = 0, colour = "white", size = 3) +
geom_line() +
ggtitle("Logistic Regression Remaining pattern")
rf %>%
add_residuals(rf_mod) %>%
ggplot(aes(year, resid)) +
geom_hline(yintercept = 0, colour = "white", size = 3) +
geom_line() +
ggtitle("Random Forest Remaining pattern")
nb %>%
add_residuals(nb_mod) %>%
ggplot(aes(year, resid)) +
geom_hline(yintercept = 0, colour = "white", size = 3) +
geom_line() +
ggtitle("Naive Bayes Remaining pattern")
dt %>%
add_residuals(dt_mod) %>%
ggplot(aes(year, resid)) +
geom_hline(yintercept = 0, colour = "white", size = 3) +
geom_line() +
ggtitle("Decision Tree Remaining pattern")
by_side <- foo %>%
group_by(classifier, sampling, technique) %>%
nest()
by_side## # A tibble: 33 × 4
## # Groups: classifier, sampling, technique [33]
## classifier sampling technique data
## <chr> <chr> <chr> <list>
## 1 Naive Bayes Imbalanced N/A <tibble [9 × 5]>
## 2 Logistic Reg Imbalanced N/A <tibble [9 × 5]>
## 3 XGBoost Imbalanced N/A <tibble [9 × 5]>
## 4 DecisionTree Imbalanced N/A <tibble [9 × 5]>
## 5 Random Forest Imbalanced N/A <tibble [1 × 5]>
## 6 Naive Bayes Undersampling NearMiss <tibble [9 × 5]>
## 7 Logistic Reg Undersampling NearMiss <tibble [9 × 5]>
## 8 XGBoost Undersampling NearMiss <tibble [9 × 5]>
## 9 DecisionTree Undersampling NearMiss <tibble [9 × 5]>
## 10 Random Forest Undersampling NearMiss <tibble [1 × 5]>
## # … with 23 more rowsby_side$data[[1]]## # A tibble: 9 × 5
## x y year lclassifier resid
## <dbl> <dbl> <int> <dbl> <dbl>
## 1 74.5 16.7 2010 -2.18 6.99
## 2 83.0 12.7 2012 -0.396 15.4
## 3 74.7 13.3 2013 -5.46 7.20
## 4 84.5 9.13 2014 -2.88 17.0
## 5 82.1 6.44 2015 -7.21 14.6
## 6 8.44 0.25 2016 -63.9 -59.1
## 7 82.3 7.92 2017 -5.62 14.8
## 8 79.1 4.85 2018 -10.9 11.6
## 9 80.1 3.79 2019 -11.2 12.6foo_model <- function(df){
lm(x ~ y, data = df)
}
models <- map(by_side$data, foo_model)
by_side <- by_side %>%
mutate(model = map(data, foo_model))
# by_side %>%
# filter(classifier == "XGBoost")
by_side %>%
arrange(classifier, sampling, technique)## # A tibble: 33 × 5
## # Groups: classifier, sampling, technique [33]
## classifier sampling technique data model
## <chr> <chr> <chr> <list> <list>
## 1 DecisionTree FS Standard Scalar <tibble [1 × 5]> <lm>
## 2 DecisionTree Imbalanced N/A <tibble [9 × 5]> <lm>
## 3 DecisionTree Oversampling ROS <tibble [9 × 5]> <lm>
## 4 DecisionTree Oversampling SMOTE <tibble [9 × 5]> <lm>
## 5 DecisionTree Undersampling NearMiss <tibble [9 × 5]> <lm>
## 6 DecisionTree Undersampling RUS <tibble [9 × 5]> <lm>
## 7 DecisionTree Undersampling Tomelinks <tibble [9 × 5]> <lm>
## 8 Logistic Reg FS SS & SKB <tibble [1 × 5]> <lm>
## 9 Logistic Reg Imbalanced N/A <tibble [9 × 5]> <lm>
## 10 Logistic Reg Oversampling ROS <tibble [9 × 5]> <lm>
## # … with 23 more rowsby_side <- by_side %>%
mutate(
resids = map2(data, model, add_residuals)
)
by_side## # A tibble: 33 × 6
## # Groups: classifier, sampling, technique [33]
## classifier sampling technique data model resids
## <chr> <chr> <chr> <list> <list> <list>
## 1 Naive Bayes Imbalanced N/A <tibble [9 × 5]> <lm> <tibble>
## 2 Logistic Reg Imbalanced N/A <tibble [9 × 5]> <lm> <tibble>
## 3 XGBoost Imbalanced N/A <tibble [9 × 5]> <lm> <tibble>
## 4 DecisionTree Imbalanced N/A <tibble [9 × 5]> <lm> <tibble>
## 5 Random Forest Imbalanced N/A <tibble [1 × 5]> <lm> <tibble>
## 6 Naive Bayes Undersampling NearMiss <tibble [9 × 5]> <lm> <tibble>
## 7 Logistic Reg Undersampling NearMiss <tibble [9 × 5]> <lm> <tibble>
## 8 XGBoost Undersampling NearMiss <tibble [9 × 5]> <lm> <tibble>
## 9 DecisionTree Undersampling NearMiss <tibble [9 × 5]> <lm> <tibble>
## 10 Random Forest Undersampling NearMiss <tibble [1 × 5]> <lm> <tibble>
## # … with 23 more rowsresids <- unnest(by_side, resids)
resids## # A tibble: 225 × 10
## # Groups: classifier, sampling, technique [33]
## classifier sampling technique data model x y year lclassifier
## <chr> <chr> <chr> <list> <lis> <dbl> <dbl> <int> <dbl>
## 1 Naive Bayes Imbalanc… N/A <tibble> <lm> 74.5 16.7 2010 -2.18
## 2 Naive Bayes Imbalanc… N/A <tibble> <lm> 83.0 12.7 2012 -0.396
## 3 Naive Bayes Imbalanc… N/A <tibble> <lm> 74.7 13.3 2013 -5.46
## 4 Naive Bayes Imbalanc… N/A <tibble> <lm> 84.5 9.13 2014 -2.88
## 5 Naive Bayes Imbalanc… N/A <tibble> <lm> 82.1 6.44 2015 -7.21
## 6 Naive Bayes Imbalanc… N/A <tibble> <lm> 8.44 0.25 2016 -63.9
## 7 Naive Bayes Imbalanc… N/A <tibble> <lm> 82.3 7.92 2017 -5.62
## 8 Naive Bayes Imbalanc… N/A <tibble> <lm> 79.1 4.85 2018 -10.9
## 9 Naive Bayes Imbalanc… N/A <tibble> <lm> 80.1 3.79 2019 -11.2
## 10 Logistic Reg Imbalanc… N/A <tibble> <lm> 72.8 21.2 2010 7.68
## # … with 215 more rows, and 1 more variable: resid <dbl>resids %>%
ggplot(aes(year, resid)) +
geom_line(aes(group = classifier), alpha = 1 / 3) +
geom_smooth(se = FALSE) +
ggtitle("Classifier Residuals")
#> `geom_smooth()` using method = 'gam' and formula 'y ~ s(x, bs = "cs")'resids %>%
ggplot(aes(year, resid)) +
geom_line(aes(group = classifier), alpha = 1 / 3) +
geom_smooth(se = FALSE) +
ggtitle("Classifier Residuals")
#> `geom_smooth()` using method = 'gam' and formula 'y ~ s(x, bs = "cs")'resids %>%
ggplot(aes(year, resid)) +
geom_line(aes(group = technique), alpha = 1 / 3) +
geom_smooth(se = FALSE) +
ggtitle("Technique Residuals")
#> `geom_smooth()` using method = 'gam' and formula 'y ~ s(x, bs = "cs")'resids %>%
ggplot(aes(year, resid, group = technique)) +
geom_line(alpha = 1 / 3) +
facet_wrap(~classifier)
resids %>%
ggplot(aes(year, resid, group = sampling)) +
geom_line(alpha = 1 / 3) +
facet_wrap(~technique)
resids %>%
ggplot(aes(year, resid, group = classifier)) +
geom_line(alpha = 1 / 3) +
facet_wrap(~sampling)
broom::glance(xg_mod)## # A tibble: 1 × 12
## r.squared adj.r.squared sigma statistic p.value df logLik AIC BIC
## <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 0.329 0.316 23.2 25.5 0.00000588 1 -245. 497. 503.
## # … with 3 more variables: deviance <dbl>, df.residual <int>, nobs <int>broom::glance(nb_mod)## # A tibble: 1 × 12
## r.squared adj.r.squared sigma statistic p.value df logLik AIC BIC
## <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 0.259 0.245 12.4 18.5 0.0000722 1 -216. 437. 443.
## # … with 3 more variables: deviance <dbl>, df.residual <int>, nobs <int>broom::glance(lr_mod)## # A tibble: 1 × 12
## r.squared adj.r.squared sigma statistic p.value df logLik AIC BIC
## <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 0.286 0.273 10.2 21.2 0.0000259 1 -205. 416. 422.
## # … with 3 more variables: deviance <dbl>, df.residual <int>, nobs <int>broom::glance(dt_mod)## # A tibble: 1 × 12
## r.squared adj.r.squared sigma statistic p.value df logLik AIC BIC
## <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 0.0641 0.0464 15.5 3.63 0.0622 1 -228. 461. 467.
## # … with 3 more variables: deviance <dbl>, df.residual <int>, nobs <int>broom::glance(rf_mod)## # A tibble: 1 × 12
## r.squared adj.r.squared sigma statistic p.value df logLik AIC BIC
## <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 0.986 0.982 2.46 273. 0.0000788 1 -12.7 31.4 30.8
## # … with 3 more variables: deviance <dbl>, df.residual <int>, nobs <int>by_side %>%
mutate(glance = map(model, broom::glance)) %>%
unnest(glance)## # A tibble: 33 × 18
## # Groups: classifier, sampling, technique [33]
## classifier sampling technique data model resids r.squared adj.r.squared
## <chr> <chr> <chr> <list> <lis> <list> <dbl> <dbl>
## 1 Naive Bay… Imbalan… N/A <tibble> <lm> <tibble> 0.270 0.166
## 2 Logistic … Imbalan… N/A <tibble> <lm> <tibble> 0.424 0.341
## 3 XGBoost Imbalan… N/A <tibble> <lm> <tibble> 0.703 0.660
## 4 DecisionT… Imbalan… N/A <tibble> <lm> <tibble> 0.746 0.710
## 5 Random Fo… Imbalan… N/A <tibble> <lm> <tibble> 0 0
## 6 Naive Bay… Undersa… NearMiss <tibble> <lm> <tibble> 0.909 0.896
## 7 Logistic … Undersa… NearMiss <tibble> <lm> <tibble> 0.182 0.0649
## 8 XGBoost Undersa… NearMiss <tibble> <lm> <tibble> 0.920 0.909
## 9 DecisionT… Undersa… NearMiss <tibble> <lm> <tibble> 0.732 0.693
## 10 Random Fo… Undersa… NearMiss <tibble> <lm> <tibble> 0 0
## # … with 23 more rows, and 10 more variables: sigma <dbl>, statistic <dbl>,
## # p.value <dbl>, df <dbl>, logLik <dbl>, AIC <dbl>, BIC <dbl>,
## # deviance <dbl>, df.residual <int>, nobs <int>glance <- by_side %>%
mutate(glance = map(model, broom::glance)) %>%
unnest(glance, .drop = TRUE)
# glance
glance %>%
arrange(r.squared)## # A tibble: 33 × 18
## # Groups: classifier, sampling, technique [33]
## classifier sampling technique data model resids r.squared adj.r.squared
## <chr> <chr> <chr> <list> <lis> <list> <dbl> <dbl>
## 1 Random Fo… Imbalan… N/A <tibble> <lm> <tibble> 0 0
## 2 Random Fo… Undersa… NearMiss <tibble> <lm> <tibble> 0 0
## 3 Random Fo… Oversam… SMOTE <tibble> <lm> <tibble> 0 0
## 4 Random Fo… Oversam… ROS <tibble> <lm> <tibble> 0 0
## 5 Random Fo… Unsersa… RUS <tibble> <lm> <tibble> 0 0
## 6 Random Fo… Undersa… Tomelinks <tibble> <lm> <tibble> 0 0
## 7 DecisionT… FS Standard… <tibble> <lm> <tibble> 0 0
## 8 Naive Bay… FS Extra Tr… <tibble> <lm> <tibble> 0 0
## 9 Logistic … FS SS & SKB <tibble> <lm> <tibble> 0 0
## 10 Logistic … Undersa… NearMiss <tibble> <lm> <tibble> 0.182 0.0649
## # … with 23 more rows, and 10 more variables: sigma <dbl>, statistic <dbl>,
## # p.value <dbl>, df <dbl>, logLik <dbl>, AIC <dbl>, BIC <dbl>,
## # deviance <dbl>, df.residual <int>, nobs <int>glance %>%
ggplot(aes(classifier, r.squared)) +
geom_jitter(width = 0.5) +
ggtitle("Classifier R.Squared")
# glance %>%
# ggplot(aes(sampling, r.squared)) +
# geom_jitter(width = 0.5) +
# ggtitle("Sampling R.Squared")
# Creating error: Validate mapping
# glance %>%
# ggplot(aes(year, r.squared)) +
# geom_jitter(width = 0.5) +
# geom_jitter("Year R.Squared")glance %>%
ggplot(aes(sampling, r.squared)) +
geom_jitter(width = 0.5) +
ggtitle("Sampling R.Squared")
bad_fit <- filter(glance, r.squared < 0.25)
bad_fit ## # A tibble: 10 × 18
## # Groups: classifier, sampling, technique [10]
## classifier sampling technique data model resids r.squared adj.r.squared
## <chr> <chr> <chr> <list> <lis> <list> <dbl> <dbl>
## 1 Random Fo… Imbalan… N/A <tibble> <lm> <tibble> 0 0
## 2 Logistic … Undersa… NearMiss <tibble> <lm> <tibble> 0.182 0.0649
## 3 Random Fo… Undersa… NearMiss <tibble> <lm> <tibble> 0 0
## 4 Random Fo… Oversam… SMOTE <tibble> <lm> <tibble> 0 0
## 5 Random Fo… Oversam… ROS <tibble> <lm> <tibble> 0 0
## 6 Random Fo… Unsersa… RUS <tibble> <lm> <tibble> 0 0
## 7 Random Fo… Undersa… Tomelinks <tibble> <lm> <tibble> 0 0
## 8 DecisionT… FS Standard… <tibble> <lm> <tibble> 0 0
## 9 Naive Bay… FS Extra Tr… <tibble> <lm> <tibble> 0 0
## 10 Logistic … FS SS & SKB <tibble> <lm> <tibble> 0 0
## # … with 10 more variables: sigma <dbl>, statistic <dbl>, p.value <dbl>,
## # df <dbl>, logLik <dbl>, AIC <dbl>, BIC <dbl>, deviance <dbl>,
## # df.residual <int>, nobs <int>foo %>%
semi_join(bad_fit, by = "classifier") %>%
ggplot(aes(year, x, colour = classifier)) +
geom_line() +
ggplot("Classifier for Precision")## Error in `fortify()`:
## ! `data` must be a data frame, or other object coercible by `fortify()`, not a character vector.foo %>%
semi_join(bad_fit, by = "classifier") %>%
ggplot(aes(year, y, colour = classifier)) +
geom_line() +
ggplot("Classifier for Recall")## Error in `fortify()`:
## ! `data` must be a data frame, or other object coercible by `fortify()`, not a character vector.